CN112700028A - Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant - Google Patents

Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant Download PDF

Info

Publication number
CN112700028A
CN112700028A CN202011390617.0A CN202011390617A CN112700028A CN 112700028 A CN112700028 A CN 112700028A CN 202011390617 A CN202011390617 A CN 202011390617A CN 112700028 A CN112700028 A CN 112700028A
Authority
CN
China
Prior art keywords
power plant
network
virtual power
parameter
space
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202011390617.0A
Other languages
Chinese (zh)
Other versions
CN112700028B (en
Inventor
周建国
郭烨
许银亮
孙宏斌
王黎明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua-Berkeley Shenzhen Institute Preparation Office
Original Assignee
Tsinghua-Berkeley Shenzhen Institute Preparation Office
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tsinghua-Berkeley Shenzhen Institute Preparation Office filed Critical Tsinghua-Berkeley Shenzhen Institute Preparation Office
Priority to CN202011390617.0A priority Critical patent/CN112700028B/en
Publication of CN112700028A publication Critical patent/CN112700028A/en
Application granted granted Critical
Publication of CN112700028B publication Critical patent/CN112700028B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Strategic Management (AREA)
  • Artificial Intelligence (AREA)
  • Human Resources & Organizations (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application discloses a virtual power plant equivalent inertia and damping space-time distribution estimation method and system. The estimation method of the equivalent inertia and damping space-time distribution of the virtual power plant comprises the following steps: obtaining a virtual power plant state parameter, and obtaining a parameter data set according to the virtual power plant state parameter; preprocessing the parameter data set to obtain a training set and a testing set; constructing a space-time distribution estimation model according to a deep neural network, and training the space-time distribution estimation model according to the training set; and obtaining a key parameter set according to the space-time distribution estimation model, and calculating the key parameter set to obtain a first probability distribution result and a second probability distribution result. The first probability distribution result can be used for representing the space-time distribution of the equivalent inertia and the damping of the distributed power supply in the virtual power plant, and the second probability distribution result can be used for representing the space-time distribution of the equivalent inertia and the damping of the virtual power plant.

Description

Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant
Technical Field
The application relates to the technical field of energy Internet, in particular to a virtual power plant equivalent inertia and damping space-time distribution estimation method and system.
Background
The renewable energy distributed power supply and the flexible load control which take the power electronic converter as a main interface are widely applied to the power system, so that the system inertia and the damping of the power system are reduced, and the safe and stable operation of the power system and the large-scale efficient integration and utilization of renewable energy sources are not facilitated. How to estimate the equivalent inertia of the virtual power plant and the space-time distribution of the damping is a key point and difficulty of the virtual power plant participating in a novel inertia auxiliary service market and providing a safe and stable frequency response support for a system.
At present, operators of the power system cannot master inertia-damping levels and distribution conditions of the inertia-damping levels in different areas in real time. The research of the inertia estimation problem mainly focuses on the estimation of the total equivalent inertia of a large system, the prediction problem is not involved, the estimation of the damping and the inertia of each area is neglected, and the coordination optimization and the formulation of a control strategy can not be carried out on system operators to form specific effective guidance.
Disclosure of Invention
The present application is directed to solving at least one of the problems in the prior art. Therefore, based on the estimation method and the estimation system for the equivalent inertia and the damping space-time distribution of the virtual power plant, the parameters of the virtual power plant are estimated through a space-time distribution estimation model, a first probability distribution result and a second probability distribution result are obtained, the first probability distribution result represents the space-time distribution of the equivalent inertia and the damping of the distributed power supply in the virtual power plant, and the second probability distribution result represents the space-time distribution of the aggregate equivalent inertia and the damping of the virtual power plant.
The embodiment of the application provides a method for estimating equivalent inertia and damping space-time distribution of a virtual power plant in a first aspect, which comprises the following steps:
acquiring virtual power plant state parameters, and acquiring a parameter data set according to the virtual power plant state parameters;
preprocessing the parameter data set to obtain a training set and a testing set;
constructing a space-time distribution estimation model according to a deep neural network, and training the space-time distribution estimation model according to the training set;
and obtaining a key parameter set according to the space-time distribution estimation model, and calculating the key parameter set to obtain a first probability distribution result and a second probability distribution result.
The estimation method for the equivalent inertia and damping space-time distribution of the virtual power plant in the embodiment of the application has the following technical effects: by accurately estimating and predicting the time-space distribution of the equivalent inertia-damping in the current and future periods, the training speed is high, the estimation precision is high, any hypothesis and constraint on the estimation target equivalent inertia-damping are not needed, the method is suitable for different operation scenes and operation conditions, the universality is strong, the robustness is good, the virtual power plant is promoted to participate in a novel inertia auxiliary service market, and a safe and stable frequency response support is provided for the system.
Further, the acquiring of the virtual power plant state parameters includes: obtaining a parameter dynamic equation according to the parameters of the synchronous distributed power supply and the renewable energy source, and obtaining the state parameters of the virtual power plant according to the parameter dynamic equation; wherein, the virtual power plant state parameters include: a voltage parameter, a current parameter, and a frequency deviation vector.
Further, the virtual power plant state parameters specifically include: the virtual power plant and power grid connection point bus voltage parameter, current parameter and frequency deviation vector, and the virtual power plant each distributed power source output bus voltage parameter, current parameter and frequency deviation vector.
Further, the preprocessing the parameter data set to obtain a training set and a testing set includes: performing hypothesis checking on the parameter data set to obtain a checking parameter data set; restoring the checking parameter data set to obtain a complete parameter data set; dividing the complete parameter data set to obtain the training set and the test set;
the method for constructing the space-time distribution estimation model according to the deep neural network comprises the following steps: the method comprises the steps of constructing a first time-space distribution estimation model and a second time-space distribution estimation model according to a deep neural network, wherein the first time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on a distributed power supply, and the second time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on virtual power plant polymerization.
Further, the deep neural network comprises a network prediction network and a distribution approximation network; the network prediction network comprises: the distributed power supply network prediction network and the virtual power plant network prediction network are characterized in that the distributed approximation network comprises: the system comprises a distributed power supply distribution approximate network and a virtual power plant aggregation distribution approximate network.
Furthermore, the distributed power supply network prediction is formed by connecting a network short-circuit residual error network, a time cycle neural network input layer network, a full-communication layer network and an output layer network in series; the distributed power supply distribution approximation network is formed by connecting a short-circuit residual error network, two full communication layers and an output layer in series; the virtual power plant network prediction network is formed by connecting a distributed power supply prediction network and an independent network prediction network in parallel; the distributed power supply distribution approximate network and the virtual power plant aggregation distribution approximate network are deep residual error networks.
Further, the obtaining a parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy parameter and the obtaining the state parameter of the virtual power plant according to the parameter dynamic equation include: obtaining a first parameter dynamic equation and a second parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy parameter; obtaining a first virtual power plant state parameter according to the first parameter dynamic equation, and obtaining a second virtual power plant state parameter according to the second parameter dynamic equation; the first parameter dynamic equation is used for representing the distributed power supply state, and the second parameter dynamic equation is used for representing the power supply state of the virtual power plant connected with the power grid.
Further, the virtual power plant is formed by connecting a distributed power supply based on a traditional synchronous generator, a distributed power supply based on renewable energy power generation and other flexible loads through a power network; the control strategy of the distributed power supply in the virtual power plant can be a traditional synchronous generator control strategy, a droop control strategy, a virtual synchronous machine control strategy and other common control strategies.
Further, the obtaining a key parameter set according to the spatio-temporal distribution estimation model includes: and obtaining a distributed key parameter set according to the first virtual power plant state parameter and a first preset index model, and obtaining a virtual power plant key parameter set according to the second virtual power plant state parameter and a second preset index model.
Further, the first parametric dynamical equation:
Figure BDA0002812413210000031
wherein, Δ ωDG,iFor output frequency deviation, M, of distributed power supplyDG,iIs equivalent inertia, D, of the distributed power supplyDG,iFor equivalent damping, K, of distributed power suppliesDG,iIs the synchronous power coefficient of the distributed power supply;
the second parametric dynamic equation:
Figure BDA0002812413210000032
wherein, Δ ωVPPFor the frequency deviation of the virtual power plant and the power grid connection point bus, MVPPIs equivalent inertia, D, of a virtual power plantVPPEquivalent damping, K, for a virtual power plantVPPIs the synchronous power coefficient of the virtual power plant.
Further, the training the spatio-temporal distribution estimation model according to the training set includes: and performing off-line training and on-line training on the space-time distribution estimation model according to the loss function and the training set.
A second aspect of the embodiments of the present application provides a system for estimating a spatial-temporal distribution of equivalent inertia and damping of a virtual power plant, including
The data acquisition and storage module is used for acquiring virtual power plant state parameters and obtaining a parameter data set according to the virtual power plant state parameters; the data preprocessing module is used for preprocessing the parameter data set to obtain a training set and a test set; the distributed power supply estimation module is used for constructing a first time-space distribution estimation model according to the deep neural network and obtaining a distributed key parameter set; the virtual power plant aggregation estimation module is used for constructing a second space-time distribution estimation model according to the deep neural network and obtaining a virtual power plant key parameter set; and the space-time distribution output module is used for outputting a first probability distribution result according to the distributed key parameter set and outputting a second probability distribution result according to the virtual power plant key parameter set.
Additional aspects and advantages of the present application will be set forth in part in the description which follows and, in part, will be obvious from the description.
Drawings
The present application is further described with reference to the following figures and examples, in which:
FIG. 1 is a schematic diagram of a virtual power plant framework of a method for estimating an equivalent inertia and damping space-time distribution of a virtual power plant according to an embodiment of the present disclosure;
FIG. 2 is a flowchart illustrating a method for estimating the equivalent inertia and damping space-time distribution of a virtual power plant according to another embodiment of the present disclosure;
FIG. 3 is a block diagram of a virtual power plant equivalent inertia and damping spatio-temporal distribution estimation model in an embodiment of the present application;
FIG. 4 is a schematic diagram of a DG-NFN network architecture according to an embodiment of the present application;
fig. 5A to fig. 5C are schematic diagrams of a short-circuited residual error network according to an embodiment of the present application;
FIG. 6 is a diagram of a virtual power plant network prediction network structure according to an embodiment of the present application;
FIG. 7 is a schematic diagram of a distributed power supply equivalent inertia and damping distribution approximation network according to an embodiment of the present application;
FIG. 8 is a schematic diagram of an approximate network structure of the virtual power plant aggregate equivalent inertia and damping distribution in an embodiment of the present application;
FIG. 9 is a block diagram of a virtual power plant equivalent inertia and damping spatio-temporal distribution estimation system according to an embodiment of the present application.
Description of reference numerals: 100. a data acquisition and storage module; 200. a data preprocessing module; 300. a distributed power source estimation module; 400. a virtual power plant aggregation estimation module; 500. and a space-time distribution output module.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application.
In the description of the present application, it is to be understood that the positional descriptions, such as the directions of up, down, front, rear, left, right, etc., referred to herein are based on the directions or positional relationships shown in the drawings, and are only for convenience of description and simplicity of description, and do not indicate or imply that the referred device or element must have a specific direction, be constructed and operated in a specific direction, and thus, should not be construed as limiting the present application.
In the description of the present application, the meaning of a plurality is one or more, the meaning of a plurality is two or more, and the meaning of more, less, more, etc. is understood as excluding the present number, and the meaning of more, less, more, etc. is understood as including the present number. If there is a description of the first and second for the purpose of distinguishing technical features, it is not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of technical features indicated.
In the description of the present application, unless otherwise expressly limited, terms such as set, mounted, connected and the like should be construed broadly, and those skilled in the art can reasonably determine the specific meaning of the terms in the present application by combining the detailed contents of the technical solutions.
In the description of the present application, reference to the description of the terms "one embodiment," "some embodiments," "an illustrative embodiment," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
In the related art, a renewable energy distributed power supply and flexible load control using a power electronic converter as a main interface are widely applied to a power system, so that system inertia and damping of the power system are reduced, and safe and stable operation of the power system and large-scale efficient integration and utilization of renewable energy are not facilitated.
At present, the research of the inertia estimation problem mainly focuses on the estimation of the total equivalent inertia of a large system, the prediction problem is not involved, the estimation of the damping and the inertia of each area is neglected, and the coordination optimization and the formulation of a control strategy can not be carried out on system operators to form specific effective guidance.
Based on the existing technical problems, the method for estimating the time-space distribution of the equivalent inertia and the damping of the virtual power plant is provided to obtain the estimation result of the cumulative probability distribution of the time-space distribution of the equivalent inertia and the damping of the distributed power supply in the virtual power plant and the aggregation equivalent inertia and the damping of the virtual power plant.
Referring to fig. 1 to 2, the virtual power plant is formed by connecting a distributed power source based on a synchronous generator, a distributed power source based on renewable energy power generation, and other flexible load power networks, and the synchronous generator is connected to an external main power network through a Point of Common Coupling (PCC).
The application provides an estimation method of equivalent inertia and damping space-time distribution of a virtual power plant, which comprises the following steps:
s100, acquiring virtual power plant state parameters, and acquiring a parameter data set according to the virtual power plant state parameters;
s200, preprocessing a parameter data set to obtain a training set and a testing set;
s300, constructing a space-time distribution estimation model according to the deep neural network, and training the space-time distribution estimation model according to a training set;
and S400, obtaining a key parameter set according to the space-time distribution estimation model, and calculating the key parameter set to obtain a first probability distribution result and a second probability distribution result.
And obtaining a parameter data set based on the virtual power plant state parameters, and obtaining a training set used for training the time-space distribution estimation model and a testing set used for testing the time-space distribution estimation model according to the parameter data set. And constructing a space-time distribution estimation model through a deep neural network to obtain a primary calculation model, and training the space-time distribution estimation model through a training set. The model is verified by grouping the parameter data sets to obtain a test set and testing the spatio-temporal distribution estimation model through the test set. And obtaining a first probability distribution result and a second probability distribution result according to the key parameter set. The first probability distribution result can be used for representing the space-time distribution of the equivalent inertia and the damping of the distributed power supply in the virtual power plant, and the second probability distribution result can be used for representing the space-time distribution of the equivalent inertia and the damping of the virtual power plant, so that specific and effective guidance is provided for system operators to carry out coordination optimization and formulation of a control strategy.
The estimation method for the equivalent inertia and damping space-time distribution of the virtual power plant, which is provided by the embodiment of the application, can accurately estimate and predict the equivalent inertia and damping space-time distribution in the current and future periods of time, has the advantages of high training speed and high estimation precision, and does not need any hypothesis and constraint on the estimated target, namely the equivalent inertia and the damping. And the space-time distribution estimation model is suitable for different operation scenes and operation conditions, has strong universality and good robustness, and has important significance and very high practical value for promoting a virtual power plant to participate in a novel inertia auxiliary service market, providing a safe and stable frequency response support for a system, promoting large-scale and high-efficiency integration of renewable energy sources and the like. Can be widely applied to the technical field of energy Internet.
In some embodiments, obtaining the virtual plant state parameter comprises: obtaining a parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy source parameter, and obtaining a virtual power plant state parameter according to the parameter dynamic equation;
wherein, virtual power plant state parameter includes: a voltage parameter, a current parameter, and a frequency deviation vector. The virtual power plant state parameters specifically include: the virtual power plant and power grid connection point bus voltage parameter, current and frequency deviation vector, each distributed power source output bus voltage parameter, current parameter and frequency deviation vector in the virtual power plant.
It can be understood that the virtual power plant state parameters obtained in step S100 are measurement data obtained by measuring the virtual power plant by the vector measurement unit or other measurement devices, a plurality of sub-parameter data sets are obtained according to the virtual power plant state parameters, and a parameter data set is established according to the relationship between the plurality of sub-parameter data sets.
Further, the virtual plant state parameters include a voltage parameter, a current parameter, and a frequency deviation vector. And obtaining a parameter set for effectively representing the virtual power plant state by taking the measured voltage parameter, the measured current parameter and the measured frequency deviation vector as the representation of the virtual power plant state.
By applying a test time period t-Deltat, t]In, measure in order to obtain virtual power plant state parameter to virtual power plant, virtual power plant state parameter specifically includes: virtual power plant and power grid connection point bus voltage Vvpp,gridA current I corresponding theretovpp,gridAnd the corresponding frequency deviation vector delta omegavpp(ii) a Output bus voltage V of each distributed power supply in virtual power plantDG,iCorresponding current IDG,iAnd the corresponding frequency deviation vector delta omegaDG,i。
The virtual power plant state parameters of the virtual power plant are effectively represented by respectively detecting the state parameters of the connection point of the virtual power plant and the power grid and the state parameters of the output buses of all the distributed power supplies in the virtual power plant, and the space-time distribution estimation model is trained according to the virtual power plant state parameters.
In some embodiments, preprocessing the parameter data set to obtain a training set and a test set, includes: performing hypothesis checking processing on the parameter data set to obtain a checking parameter data set; restoring the checking parameter data set to obtain a complete parameter data set; and dividing the complete parameter data set to obtain a training set and a test set.
And performing hypothesis checking on the parameter data set by using Bayesian hypothesis test to check and check the parameter data set so as to obtain an integrity result corresponding to the parameter data set. And restoring the correction parameter data set according to the integrity result to obtain an integral parameter data set. A plurality of complete parameter data sets are randomly transformed into two groups of data sets so as to achieve a training set and a testing set. The training set comprises NtrA complete parameter data set, the test set including NteThe complete parameter data set. And training a space-time distribution estimation model constructed according to the deep neural network through a training set.
In some embodiments, constructing the spatio-temporal distribution estimation model from the deep neural network comprises: and constructing a first time-space distribution estimation model and a second time-space distribution estimation model according to the deep neural network, wherein the first time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on the distributed power supply, and the second time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on the virtual power plant aggregation.
And respectively constructing a first time-space distribution estimation model and a second time-space distribution estimation model according to the deep neural network so as to perform equivalent inertia damping estimation on the aggregation of the distributed power supply and the virtual power plant. The first time-space distribution estimation model is an equivalent inertia-damping estimation model of a distributed power supply in the virtual power plant, and the second time-space distribution estimation model is a virtual power plant polymerization equivalent inertia-damping estimation model.
Referring to fig. 3, in some embodiments, the deep neural network includes a network prediction network and a distribution approximation network; the network prediction network comprises: a distributed power network prediction network (DG-NFN) and a virtual power plant network prediction network (VPP-NFN), the distribution approximation network comprising: distributed-power-distribution-approximation network (DG-EID-DAN) and virtual-plant-aggregation-distribution-approximation network (VPP-EID-DAN).
It can be understood that a space-time distribution estimation model is constructed through a network prediction network (NFN) and a Distribution Approximation Network (DAN) to estimate and calculate distribution approximation according to space-time distribution of equivalent inertia and damping.
Furthermore, the network prediction network and the distribution approximation network are used as a deep neural network for constructing a space-time distribution estimation model, so that the space-time distribution estimation model can perform network prediction and distribution approximation processing on the state of the virtual power plant.
Further, virtual power plant state parameters of the distributed power supply and the virtual power plant are respectively predicted through a distributed power supply network prediction network (DG-NFN) and a virtual power plant network prediction network (VPP-NFN); and carrying out approximate estimation on virtual power plant state parameters of the distributed power supply and the virtual power plant through a distributed power supply distribution approximate network (DG-EID-DAN) and a virtual power plant aggregation distribution approximate network (VPP-EID-DAN).
Referring to fig. 3 and 4, in some embodiments, the distributed power network prediction network is formed by connecting a short-circuit residual error network, a time-cycle neural network input layer network, a full-connectivity layer network, and an output layer network in series; the distributed power distribution approximation network is formed by connecting a short-circuit residual error network, two full communication layers and an output layer in series; the virtual power plant network prediction network is formed by connecting a distributed power supply prediction network and an independent network prediction network in parallel; the distributed power source distribution approximate network and the virtual power plant aggregation distribution approximate network are deep residual error networks.
Referring to fig. 5A to 5C together, for example, the distributed power network prediction network (DG-NFN) is formed by connecting Q blocks of short-circuited residual error networks with width M, a time-cycled neural network input layer network (LSTM input layer network) with width M, a full-pass layer network with width L, and an output layer network in series.
The distributed power supply distribution approximation network is a distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN) and is used for carrying out approximation calculation on equivalent inertia and damping distribution. The distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN) is formed by connecting K short-circuit residual error networks with the width of L based on a full-communication layer, two full-communication layers with the same width of L and an output layer in series. The short-circuit residual error network structure is the same as the residual error network structure in the distributed power supply network prediction network.
The virtual power plant network prediction network (VPP-NFN) is formed by connecting a distributed power supply prediction network (all distributed power supply network prediction networks of distributed power supplies in a virtual power plant) and an independent network prediction network (S-NFN) in parallel. The virtual power plant aggregation distribution approximation network (virtual power plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN)) and the distributed power supply distribution approximation network (distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN)) are composed of similar depth residual error networks. Each full communication layer of the virtual plant aggregate distribution approximation network has a width of (N)DG+1)·L，NDGThe number of distributed power sources in the virtual power plant.
In some embodiments, obtaining a parameter dynamic equation according to the synchronous distributed power source parameter and the renewable energy source parameter, and obtaining a virtual power plant state parameter according to the parameter dynamic equation includes: obtaining a first parameter dynamic equation and a second parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy source parameter; obtaining a first virtual power plant state parameter according to a first parameter dynamic equation, and obtaining a second virtual power plant state parameter according to a second parameter dynamic equation; the first parameter dynamic equation is used for representing the distributed power supply state, and the second parameter dynamic equation is used for representing the power supply state of the virtual power plant connected with the power grid.
And obtaining a first parameter dynamic equation for representing the state of the distributed power supply and a second parameter dynamic equation for representing the state of the power supply connected with the power grid of the virtual power plant according to the synchronous distributed power supply parameter and the renewable energy parameter. And quickly solving the key parameter set through the first parameter dynamic equation and the second parameter dynamic equation.
And representing the state of the distributed power supply through a first parameter dynamic equation, and representing the state of the power supply connected with the virtual power plant and the power grid through a second parameter dynamic equation, so that the dynamic relations of all parameters in the distributed power supply, the virtual power plant and the power supply connected with the power grid are dynamically represented respectively.
In some embodiments, deriving the set of key parameters from the spatio-temporal distribution estimation model comprises: and obtaining a distributed key parameter set according to the first virtual power plant state parameter and the first preset index model, and obtaining a virtual power plant key parameter set according to the second virtual power plant state parameter and the second preset index model.
Obtaining a distributed key parameter set according to the first virtual power plant state parameter and the first preset index model, and obtaining a first probability distribution result according to the distributed key parameter set; and obtaining a virtual power plant key parameter set according to the second virtual power plant state parameter and the second preset index model, and obtaining a second probability distribution result according to the virtual power plant key parameter set. And respectively obtaining a first probability distribution result and a second probability distribution result to obtain the space-time distribution of the equivalent inertia and the damping.
In some embodiments, the first parametric dynamical equation:
Figure BDA0002812413210000091
wherein, Δ ωDG,iFor output frequency deviation, M, of distributed power supplyDG,iIs equivalent inertia, D, of the distributed power supplyDG,iFor equivalent damping, K, of distributed power suppliesDG,iIs the synchronous power coefficient of the distributed power supply;
second parametric dynamical equation:
Figure BDA0002812413210000092
wherein, Δ ωVPPOutput frequency deviation, M, for virtual power plant to grid connection pointVPPIs equivalent inertia, D, of a virtual power plantVPPEquivalent damping, K, for a virtual power plantVPPIs the synchronous power coefficient of the virtual power plant.
In some embodiments, training the spatio-temporal distribution estimation model according to a training set comprises: and performing off-line training and on-line training on the space-time distribution estimation model according to the loss function and the training set.
And performing off-line training and on-line training on the space-time distribution estimation model through a preset loss function and a training set.
The following description is made with reference to a specific calculation method.
Obtaining a parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy source parameter, and obtaining a virtual power plant state parameter according to the parameter dynamic equation; wherein, virtual power plant state parameter includes: a voltage parameter, a current parameter, and a frequency deviation vector.
For example, the virtual plant state parameters may specifically include: virtual power plant and power grid connection point bus voltage Vvpp,gridCurrent I corresponding theretovpp,gridAnd the corresponding frequency deviation vector delta omegavpp(ii) a Output bus voltage V of each distributed power supply in virtual power plantDG,iA current I corresponding theretoDG,iAnd the corresponding frequency deviation vector delta omegaDG,i。
Wherein the virtual plant state parameters may be described as:
Vvpp,grid＝[Vvpp,grid,t-Δt,Vvpp,grid,t-Δt+1,…,Vvpp,grid,t-1,Vvpp,grid,t]
Ivpp,grid＝[Ivpp,grid,t-Δt,Ivpp,grid,t-Δt+1,…,Ivpp,grid,t-1,Ivpp,grid,t]
Δωvpp＝[Δωvpp,t-Δt,Δωvpp,t-Δt+1,…,Δωvpp,t-1,Δωvpp,t] (1)
VDG,i＝[VDG,i,t-Δt,VDG,i,t-Δt+1,…,VDG,i,t-1,VDG,i,t]
IDG,i＝[IDG,i,t-Δt,IDG,i,t-Δt+1,…,IDG,i,t-1,IDG,i,t]
ΔωDG,i＝[ΔωDG,i,t-Δt,ΔωDG,i,t-Δt+1,…,ΔωDG,i,t-1,IDG,i,t] (2)
further, a first parameter dynamic equation and a second parameter dynamic equation are obtained according to the synchronous distributed power supply parameters and the renewable energy source parameters; obtaining a first virtual power plant state parameter according to a first parameter dynamic equation, and obtaining a second virtual power plant state parameter according to a second parameter dynamic equation; the first parameter dynamic equation is used for representing the distributed power supply state, and the second parameter dynamic equation is used for representing the power supply state of the virtual power plant connected with the power grid.
Further, the virtual power plant is formed by connecting a distributed power supply based on a traditional synchronous generator, a distributed power supply based on renewable energy power generation and other flexible loads through a power network; the control strategy of the distributed power supply in the virtual power plant can be a traditional synchronous generator control strategy, a droop control strategy, a virtual synchronous generator control strategy and other common control strategies.
In this embodiment, the first parameter dynamic equation is used to dynamically describe the output frequency deviation of the distributed power source in the virtual power plant, and the first parameter dynamic equation specifically includes:
Figure BDA0002812413210000101
wherein, Δ ωDG,iFor the output frequency deviation vector, M, of the distributed power supplyDG,iIs equivalent inertia, D, of the distributed power supplyDG,iFor equivalent damping, K, of distributed power suppliesDG,iIs the synchronous power coefficient of the distributed power supply. By applying a first parameter dynamic equation pair MDG,iAnd DDG,iA calculation is performed to obtain the target parameter. It will be appreciated that MDG,iAnd DDG,iAnd estimating a target for equivalent inertia and damping of the distributed power supply.
The synchronous power coefficient of the distributed power supply is calculated by the following formula (4):
Figure BDA0002812413210000102
wherein, PDG,iThe output active power of the distributed power supply can be obtained by calculating the voltage and the current obtained by measurement; deltaDG,iIs the virtual power angle of the distributed power supply, and the deviation amount deltaDG,iDeviation from frequency Δ ωDG,iThe relationship of (1) is:
Figure BDA0002812413210000103
general solution of distributed power output frequency deviation dynamic can be obtained by formula (3)
Figure BDA0002812413210000104
Wherein [ A ]DG,i,αDG,i,βDG,i,φDG,i]Is a part of the key parameter set for representing the estimation result of the distributed power space-time distribution in the embodiment of the application and comprises
Figure BDA0002812413210000111
Figure BDA0002812413210000112
Figure BDA0002812413210000113
Figure BDA0002812413210000114
Further, a second parameter dynamic equation is used for dynamically describing the frequency deviation of the virtual power plant and the power grid connection point bus, and the second parameter dynamic equation specifically comprises the following steps:
Figure BDA0002812413210000115
wherein, Δ ωVPPFor the frequency deviation of the virtual power plant and the power grid connection point bus, MVPPIs equivalent inertia, D, of a virtual power plantVPPEquivalent damping, K, for a virtual power plantVPPIs the synchronous power coefficient of the virtual power plant. MVPPAnd DVPPAnd estimating a target for equivalent inertia and damping of the virtual power plant.
Further, the synchronous power coefficient of the virtual power plant is calculated by the following formula (8):
Figure BDA0002812413210000116
wherein, PVPPThe active power exchanged between the virtual power plant and the power grid can be obtained by calculating the bus voltage and current of the virtual power plant and the power grid connection point obtained through measurement; deltaVPPIs a virtual power angle of a virtual power plant and the deviation delta thereofVPPDeviation from frequency Δ ωVPPThe relationship of (1) is:
Figure BDA0002812413210000117
general solution of virtual power plant output frequency deviation dynamics can be obtained from equation (7)
Figure BDA0002812413210000118
Wherein [ A ]VPP,αVPP,βVPP,φVPP]Part of a key parameter set for representing the estimation result of the space-time distribution of the virtual power plant in the embodiment of the application, and the key parameter set comprises
Figure BDA0002812413210000121
Figure BDA0002812413210000122
Figure BDA0002812413210000123
Figure BDA0002812413210000124
By estimating [ alpha ] in the key parameter setDG,i,βDG,i]And [ alpha ]VPP,βVPP]And combining the formulas (3), (5), (6) and (7), (9) and (10) to obtain a first probability distribution result (M)DG,i，DDG,i) Second probability distribution result (M)VPP，DVPP)。
In some embodiments, the method for constructing the spatio-temporal distribution estimation model according to the deep neural network specifically comprises the following steps: by [ alpha ] in the set of key parameters to be estimatedDG,i,βDG,i]And [ alpha ]VPP,βVPP]And (3) carrying out probability density description, wherein the corresponding probability density functions are respectively as follows:
Figure BDA0002812413210000125
Figure BDA0002812413210000126
Figure BDA0002812413210000127
Figure BDA0002812413210000128
where ρ isα,DG,i,0(·)，ρβ,DG,i,0(·)，ρα,VPP,0(. and ρ)β,VPP,0(. cndot.) is a Dirac delta function (delta function for short),
Figure BDA0002812413210000129
and
Figure BDA00028124132100001210
respectively, the probability density function over (0, 1).
Figure BDA00028124132100001211
As another part of the key parameter set corresponding to the distributed power source,
Figure BDA00028124132100001212
is another part of the set of key parameters corresponding to the virtual power plant.
Further, by [ α ] in the set of key parameters to be estimatedDG,i,βDG,i]And [ alpha ]VPP,βVPP]The cumulative probability is described, and the corresponding cumulative probability distribution functions are respectively expressed by the following formula (13) and formula (14):
Figure BDA0002812413210000131
Figure BDA0002812413210000132
Figure BDA0002812413210000133
Figure BDA0002812413210000134
wherein the content of the first and second substances,
Figure BDA0002812413210000135
further, the estimation of the key parameters is described separately.
Wherein the content of the first and second substances,
Figure BDA0002812413210000136
are described as formula (15) and formula (16), respectively, specifically:
Figure RE-GDA0002983880370000131
Figure RE-GDA0002983880370000132
Figure RE-GDA0002983880370000133
Figure RE-GDA0002983880370000134
Figure RE-GDA0002983880370000135
Figure RE-GDA0002983880370000136
Figure RE-GDA0002983880370000137
Figure RE-GDA0002983880370000138
since the network prediction network includes: the distributed power network prediction network (DG-NFN) is constructed in the following way:
the input data of the distributed power supply network prediction network are set as follows: xDG,i,t＝[VDG,i,IDG,i,ΔωDG,i]The output data is set to include: n isα,DG,i,0，nα,DG,i,1，nβ,DG,i,0，nβ,DG,i,1，WDG,i，BDG,i，WVPP,i，BVPP,i。
Wherein n isα,DG,i,0，nα,DG,i,1，nβ,DG,i,1And nβ,DG,i,0Using the hard sigmoid function sigmah(x) As an activation function, the following is specifically described:
Figure BDA0002812413210000141
WDG,iand WVPP,iThe SoftPlus function is adopted as an activation function, and the specific description is as follows:
sp(x)＝ln(1+ex) (18)
BDG,iand BVPP,iA linear activation function is used. Input data X is established by a distributed power supply network prediction networkDG,i,tAnd output data [ n ]α,DG,i,0,nα,DG,i,1,nβ,DG,i,0,nβ,DG,i,1,WDG,i,BDG,i,WVPP,i,BVPP,i]Mapping relationship between
Wherein, WDG,iAnd BDG,iAs the weight and offset of the distributed power supply distribution approximation network (distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN)); wVPP,iAnd BVPP,iAs part of the weights and offsets of a virtual plant aggregation distribution approximation network (virtual plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN)).
The distributed power supply network prediction network (DG-NFN) is formed by serially connecting a short-circuit residual error network with the width of M of a Q block, an LSTM input network with the width of M, a full-communication network with the width of L and an output network.
Further, the relationship of input x and output y of the residual network can be described as
y＝ΓR(x；[W*],[b*])+x (19)
Wherein, gamma isR(x；[W*],[b*]) Input-output mapping relation W representing shorted layer*And b*Respectively representing weight vectors and bias vectors of the neural network. The short-circuited layer in the distributed power network prediction network (DG-NFN) adopts a long-short-term memory (LSTM) network, and the internal information flow can be described as
iτ＝σ(wix·xτ+wih·hτ-1+wic·cτ-1+bi)
fτ＝σ(wfx·xτ+wfh·hτ-1+wfc·cτ-1+bf)
Figure BDA0002812413210000142
oτ＝σ(wox·xτ+woh·hτ-1+woc·cτ+bo)
Figure BDA0002812413210000143
Wherein iτTo the input gate, fτTo forget the door oτTo the output gate, cτIs a memory cell, hτThe hidden state is an implicit state with M elements, namely a weight and an offset. The information flow is further described as from xτTo hτA compact form of the mapping of (t-d + 1.., t): [ h ] oft-d+1,...,ht]＝ΓL([xt-d+1,...,xt]；[W*],[b*])。
Further, the network prediction network further comprises: the virtual power plant network prediction network (VPP-NFN) is constructed in the following way: the virtual power plant network prediction network (VPP-NFN) is formed by connecting all network prediction networks DG-NFN of distributed power supplies in a virtual power plant in parallel to form an independent network prediction network (S-NFN).
The structure of the independent network prediction network (S-NFN) is identical to that of the distributed power network prediction network (DG-NFN).
The input data of the independent network prediction network (S-NFN) is XS,t＝[Vvpp,grid,Ivpp,grid,Δωvpp]The output data is: n isα,VPP,0，nα,VPP,1，nβ,VPP,0，nβ,VPP,1，WS,VPP，BS,VPP. In the independent network prediction network (S-NFN), nα,VPP,0， nα,VPP,1，nβ,VPP,0，nβ,VPP,1Adopting the same hard sigmoid function as that in a distributed power network prediction network (DG-NFN) as an activation function, WS,VPPThe same SoftPlus function as in the distributed power network prediction network (DG-NFN) is also used as the activation function, BS,VPPA linear activation function is used.
The independent network prediction network (S-NFN) establishes XS,tAnd [ nα,VPP,0,nα,VPP,1,nβ,VPP,0,nβ,VPP,1,WS,VPP,BS,VPP]
The mapping relationship between them. WS,VPPAnd BS,VPPAs another part of the weights and offsets of the virtual plant aggregation distribution approximation network (virtual plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN)).
The deep neural network for establishing the space-time distribution estimation model further comprises a distribution approximation network, wherein the distribution approximation network comprises a distributed power supply distribution approximation network and a virtual power plant aggregation distribution approximation network.
Referring to fig. 6 to 7, the distributed power distribution approximation network is a distributed power equivalent inertia and damping distribution approximation network (DG-EID-DAN), and the virtual power plant aggregation distribution approximation network is a virtual power plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN).
The distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN) is formed by connecting K short-circuit residual error networks with the width of L based on a full-communication layer, two full-communication layers with the same width of L and an output layer in series. Wherein, the short-circuit residual error network structure in the network is the same as the residual error network structure of the distributed power supply network prediction network (DG-NFN)
The hidden layer of a distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN) adopts a sigmoid function as an activation function, the output layer adopts a linear activation function, and a W function is adoptedDG,iAnd BDG,iAs weights and offsets.
The input-output mapping relationship can be described as a deterministic function: gamma-shapedDG,i,αβ,D(·；WDG,i,BDG,i)。
Referring to fig. 8, in an embodiment, the virtual plant aggregation distribution approximation network is a virtual plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN), and the virtual plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN) is composed of a depth residual error network similar to the distributed power supply equivalent inertia and damping distribution approximation network (DG-EID-DAN), and is mainly different in that each full connectivity layer of the virtual plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN) has a width (N)DG+1)·L，NDGThe number of distributed power sources in the virtual power plant. The virtual power plant aggregation equivalent inertia and damping distribution approximation network (VPP-EID-DAN) adopts WS,VPP，BS,VPP，WVPP,iAnd BVPP,i(i＝1,…,NDG) As the weight and the offset, the input-output mapping relationship can be described as a deterministic function: gamma-shapedVPP,αβ,D(·；WVPP,BVPP)。
Wherein the content of the first and second substances,
Figure BDA0002812413210000161
specifically, the building of the spatio-temporal distribution estimation model according to the deep neural network (the network prediction network and the distribution approximation network) comprises the following steps:
for gammaDG,i,αβ,D(·；WDG,i,BDG,i) And ΓVPP,αβ,D(·；WVPP,BVPP) Carrying out normalization treatment to obtain
Figure BDA0002812413210000162
And
Figure BDA0002812413210000163
the method specifically comprises the following steps:
Figure BDA0002812413210000164
Figure BDA0002812413210000165
wherein, U·And L·Upper and lower bounds of · respectively.
Then, [ alpha ]DG,i,βDG,i]And [ alpha ]VPP,βVPP]Respectively as a cumulative probability distribution function of the estimated measures of
Figure BDA0002812413210000166
Figure BDA0002812413210000171
Wherein the content of the first and second substances,
Figure BDA0002812413210000172
further, the space-time distribution estimation model is subjected to off-line training and on-line training according to the loss function and the training set, and the method specifically comprises the following steps: and training the network prediction network through a loss function.
Wherein, the loss function adopted by the training is as follows:
Figure BDA0002812413210000173
Figure BDA0002812413210000174
wherein the content of the first and second substances,
Figure BDA0002812413210000175
is composed of
Figure BDA0002812413210000176
A derivative of (a);
Figure BDA0002812413210000177
is composed of
Figure BDA0002812413210000178
A derivative of (a);
Figure BDA0002812413210000179
and
Figure BDA00028124132100001710
are each yDG,i,t+kAnd yVPP,t+kThe realized value of (a).
In addition, a key parameter set is obtained according to the space-time distribution estimation model, and a first probability distribution result (the result of the distribution of equivalent inertia and damping cumulative probability of each distributed power supply in the virtual power plant) and a second probability distribution result (the result of the distribution of the virtual power plant aggregate equivalent inertia and damping cumulative probability) are obtained by calculating the key parameter set, which comprises the following steps:
obtaining parameters in a key parameter set according to a space-time distribution estimation model
Figure BDA0002812413210000181
And the following calculations were performed.
Figure BDA0002812413210000182
Figure BDA0002812413210000183
Figure BDA0002812413210000184
Figure BDA0002812413210000185
And reconstructing the solution of the frequency dynamic response according to the formula (5) and the formula (9) to obtain
Figure BDA0002812413210000186
And
Figure BDA0002812413210000187
the reconstruction formula is shown in the drawing (29) and the formula (30), and specifically includes:
Figure BDA0002812413210000188
Figure BDA0002812413210000189
further derivation to obtain
Figure BDA00028124132100001810
And
Figure BDA00028124132100001811
the method specifically comprises the following steps:
Figure BDA00028124132100001812
Figure BDA00028124132100001813
further, the calculation results in
Figure BDA00028124132100001814
And
Figure BDA00028124132100001815
maximum value of
Figure BDA00028124132100001816
And
Figure BDA00028124132100001817
tris composed of
Figure BDA00028124132100001818
And
Figure BDA00028124132100001819
the time at which the maximum value is taken.
In addition, the calculating the key parameter set to obtain a first probability distribution result and a second probability distribution result further includes:
Figure BDA00028124132100001820
Figure BDA00028124132100001821
Figure BDA00028124132100001822
Figure BDA00028124132100001823
wherein M isDG,iAnd DDG,iEstimation of equivalent inertia and damping for distributed power suppliesTarget of measurement, MVPPAnd DVPPAnd estimating a target for equivalent inertia and damping of the virtual power plant to represent the space-time distribution of the equivalent inertia and the damping.
Referring to fig. 9, the present application further provides a virtual power plant equivalent inertia and damping space-time distribution estimation system, which includes a data obtaining and storing module 100, configured to obtain a virtual power plant state parameter, and obtain a parameter data set according to the virtual power plant state parameter; the data preprocessing module 200 is used for preprocessing the parameter data set to obtain a training set and a test set; the distributed power estimation module 300 is used for constructing a first time-space distribution estimation model according to the deep neural network and obtaining a distributed key parameter set; the virtual power plant aggregation estimation module 400 is used for constructing a second space-time distribution estimation model according to the deep neural network and obtaining a virtual power plant key parameter set; and the space-time distribution output module 500 is configured to output a first probability distribution result according to the distributed key parameter set, and output a second probability distribution result according to the virtual power plant key parameter set.
Further, the virtual power plant equivalent inertia and damping space-time distribution estimation system performs equivalent inertia and damping space-time distribution estimation on the virtual power plant by executing the virtual power plant equivalent inertia and damping space-time distribution estimation method in the above embodiment. The distributed power estimation module 300 can estimate a key parameter set of a distributed power source in a virtual power plant and reconstruct the frequency response dynamics of the output bus of the distributed power source; the virtual plant aggregation estimation module 400 may estimate a set of key parameters for the virtual plant and reconstruct the virtual plant frequency response dynamics.
According to the estimation method and the estimation system for the equivalent inertia and the damping space-time distribution of the virtual power plant, provided by the embodiment of the application, a data set is preprocessed by acquiring the required measurement data of the distributed power supply in the virtual power plant and the connection point of the virtual power plant and a power grid; further constructing an NFN-DAN estimation model of virtual power plant equivalent inertia-damping space-time distribution based on a deep neural network, and training the deep neural network estimation model by adopting a method of combining offline and online to obtain a key parameter set required by estimation of distributed power supply equivalent inertia-damping in the virtual power plant and virtual power plant aggregate equivalent inertia-damping; and finally, obtaining the accumulative probability distribution estimation result of the time-space distribution of the equivalent inertia-damping of the distributed power supply in the virtual power plant and the aggregation equivalent inertia-damping of the virtual power plant through an equivalent inertia-damping time-space distribution output module. The estimation method and the system for the equivalent inertia-damping space-time distribution of the virtual power plant can accurately estimate and predict the equivalent inertia-damping space-time distribution in the current and future periods, are high in training speed and estimation precision, do not need any hypothesis and constraint on an estimation target, namely equivalent inertia-damping, are suitable for different operation scenes and operation conditions, are high in universality and robustness, and have important significance and high practical value in promoting the virtual power plant to participate in a novel inertia auxiliary service market, providing a safe and stable frequency response support for the system, promoting large-scale and efficient integration of renewable energy sources and the like. Can be widely applied to the technical field of energy Internet.
The embodiments of the present application have been described in detail with reference to the drawings, but the present application is not limited to the embodiments described above, and various changes can be made without departing from the spirit of the present application within the knowledge of those skilled in the art. Furthermore, the embodiments and features of the embodiments of the present application may be combined with each other without conflict.

Claims (10)

1. The estimation method of the equivalent inertia and damping space-time distribution of the virtual power plant is characterized by comprising the following steps:
acquiring virtual power plant state parameters, and acquiring a parameter data set according to the virtual power plant state parameters;
preprocessing the parameter data set to obtain a training set and a testing set;
constructing a space-time distribution estimation model according to a deep neural network, and training the space-time distribution estimation model according to the training set;
and obtaining a key parameter set according to the space-time distribution estimation model, and calculating the key parameter set to obtain a first probability distribution result and a second probability distribution result.
2. The estimation method of virtual plant equivalent inertia and damping space-time distribution according to claim 1, wherein the obtaining of the virtual plant state parameters comprises: obtaining a parameter dynamic equation according to the parameters of the synchronous distributed power supply and the renewable energy source, and obtaining the state parameters of the virtual power plant according to the parameter dynamic equation;
wherein, the virtual power plant state parameters include: a voltage parameter, a current parameter, and a frequency deviation vector.
3. The virtual power plant equivalent inertia and damping space-time distribution estimation method according to claim 2, wherein the preprocessing the parameter data set to obtain a training set and a testing set comprises:
performing hypothesis checking on the parameter data set to obtain a checking parameter data set;
restoring the checking parameter data set to obtain a complete parameter data set;
dividing the complete parameter data set to obtain the training set and the test set;
the method for constructing the space-time distribution estimation model according to the deep neural network comprises the following steps:
and constructing a first time-space distribution estimation model and a second time-space distribution estimation model according to the deep neural network, wherein the first time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on the distributed power supply, and the second time-space distribution estimation model is used for carrying out equivalent inertia damping estimation on the virtual power plant polymerization.
4. The virtual power plant equivalent inertia and damping space-time distribution estimation method according to claim 1, wherein the deep neural network comprises a network prediction network and a distribution approximation network;
the network prediction network comprises: the distributed power supply network prediction network and the virtual power plant network prediction network are characterized in that the distributed approximation network comprises: the distributed power supply distribution approximation network and the virtual power plant aggregation distribution approximation network.
5. The estimation method of equivalent inertia and damping space-time distribution of a virtual power plant according to claim 4, wherein the distributed power network prediction is formed by connecting a network short-circuit residual error network, a time cycle neural network input layer network, a full-communication layer network and an output layer network in series;
the distributed power supply distribution approximation network is formed by connecting a short-circuit residual error network, two full communication layers and an output layer in series;
the virtual power plant network prediction network is formed by connecting a distributed power supply prediction network and an independent network prediction network in parallel;
the distributed power supply distribution approximate network and the virtual power plant aggregation distribution approximate network are respectively provided with a depth residual error network with the same structure.
6. The estimation method of equivalent inertia and damping space-time distribution of a virtual power plant according to claim 2, wherein the obtaining of the parameter dynamic equation according to the synchronous distributed power parameters and the renewable energy parameters and the obtaining of the state parameters of the virtual power plant according to the parameter dynamic equation comprise:
obtaining a first parameter dynamic equation and a second parameter dynamic equation according to the synchronous distributed power supply parameter and the renewable energy source parameter;
obtaining a first virtual power plant state parameter according to the first parameter dynamic equation, and obtaining a second virtual power plant state parameter according to the second parameter dynamic equation;
the first parameter dynamic equation is used for representing the distributed power supply state, and the second parameter dynamic equation is used for representing the power supply state of the virtual power plant connected with the power grid.
7. The virtual power plant equivalent inertia and damping space-time distribution estimation method according to claim 6, wherein obtaining the key parameter set according to the space-time distribution estimation model comprises:
and obtaining a distributed key parameter set according to the first virtual power plant state parameter and a first preset index model, and obtaining a virtual power plant key parameter set according to the second virtual power plant state parameter and a second preset index model.
8. The virtual power plant equivalent inertia and damping space-time distribution estimation method according to claim 6, wherein the first parameter dynamic equation:
Figure FDA0002812413200000021
wherein, Δ ωDG,iFor output frequency deviation, M, of distributed power supplyDG,iIs equivalent inertia, D, of the distributed power supplyDG,iFor equivalent damping, K, of distributed power suppliesDG,iIs the synchronous power coefficient of the distributed power supply;
the second parametric dynamic equation:
Figure FDA0002812413200000022
wherein, Δ ωVPPFor the frequency deviation of the virtual power plant and the power grid connection point bus, MVPPIs equivalent inertia, D, of a virtual power plantVPPEquivalent damping, K, for a virtual power plantVPPIs the synchronous power coefficient of the virtual power plant.
9. The method for estimating virtual plant equivalent inertia and damping spatio-temporal distribution according to any one of claims 1 to 8, wherein the training the spatio-temporal distribution estimation model according to the training set comprises:
and performing off-line training and on-line training on the space-time distribution estimation model according to the loss function and the training set.
10. The system for estimating the space-time distribution of equivalent inertia and damping of a virtual power plant is characterized by comprising
The data acquisition and storage module is used for acquiring virtual power plant state parameters and obtaining a parameter data set according to the virtual power plant state parameters;
the data preprocessing module is used for preprocessing the parameter data set to obtain a training set and a test set;
the distributed power estimation module is used for constructing a first time-space distribution estimation model according to the deep neural network and obtaining a distributed key parameter set;
the virtual power plant aggregation estimation module is used for constructing a second space-time distribution estimation model according to the deep neural network and obtaining a virtual power plant key parameter set;
and the space-time distribution output module is used for outputting a first probability distribution result according to the distributed key parameter set and outputting a second probability distribution result according to the virtual power plant key parameter set.
CN202011390617.0A 2020-12-02 2020-12-02 Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant Active CN112700028B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011390617.0A CN112700028B (en) 2020-12-02 2020-12-02 Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011390617.0A CN112700028B (en) 2020-12-02 2020-12-02 Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant

Publications (2)

Publication Number Publication Date
CN112700028A true CN112700028A (en) 2021-04-23
CN112700028B CN112700028B (en) 2023-01-10

Family

ID=75506161

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011390617.0A Active CN112700028B (en) 2020-12-02 2020-12-02 Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant

Country Status (1)

Country Link
CN (1) CN112700028B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140277808A1 (en) * 2013-03-15 2014-09-18 Open Access Technology International, Inc. Use of Demand Response (DR) and Distributed Energy Resources (DER) to mitigate the impact of Variable Energy Resources (VER) in Power System Operation
CN110266021A (en) * 2019-05-08 2019-09-20 上海电力学院 The double adaptive dynamic frequency control methods of dimension of micro-capacitance sensor based on the virtual inertia of DFIG
CN111179110A (en) * 2019-12-06 2020-05-19 清华-伯克利深圳学院筹备办公室 Virtual power plant variable order aggregation equivalent robust dynamic model modeling method and device
CN111667109A (en) * 2020-05-29 2020-09-15 国网冀北电力有限公司计量中心 Output control method and device of virtual power plant

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140277808A1 (en) * 2013-03-15 2014-09-18 Open Access Technology International, Inc. Use of Demand Response (DR) and Distributed Energy Resources (DER) to mitigate the impact of Variable Energy Resources (VER) in Power System Operation
CN110266021A (en) * 2019-05-08 2019-09-20 上海电力学院 The double adaptive dynamic frequency control methods of dimension of micro-capacitance sensor based on the virtual inertia of DFIG
CN111179110A (en) * 2019-12-06 2020-05-19 清华-伯克利深圳学院筹备办公室 Virtual power plant variable order aggregation equivalent robust dynamic model modeling method and device
CN111667109A (en) * 2020-05-29 2020-09-15 国网冀北电力有限公司计量中心 Output control method and device of virtual power plant

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ZHONGKAI YI ET. AL: ""Bi-Level Programming for Optimal Operation of an Active Distribution Network With Multiple Virtual Power "", 《IEEE TRANSACTIONS ON SUSTAINABLE ENERGY》 *
刘方蕾等: "基于PMU同步测量的分区惯量估计方法", 《华北电力大学学报(自然科学版)》 *
季阳: ""基于多代理系统的虚拟发电厂技术及其在智能电网中的应用研究"", 《中国优秀硕士论文全文数据库 工程科技Ⅱ辑》 *

Also Published As

Publication number Publication date
CN112700028B (en) 2023-01-10

Similar Documents

Publication Publication Date Title
Bian et al. Stacked bidirectional long short-term memory networks for state-of-charge estimation of lithium-ion batteries
Chemali et al. State-of-charge estimation of Li-ion batteries using deep neural networks: A machine learning approach
US9658260B2 (en) Power system state estimation using a two-level solution
Zhao et al. A deep reinforcement learning framework for optimizing fuel economy of hybrid electric vehicles
CN105378496B (en) Estimation unit and method of estimation
Lee et al. State-of-charge estimation for electric scooters by using learning mechanisms
CN102569922A (en) Improved storage battery SOC estimation method based on consistency of unit cell
CN111142027A (en) Lithium iron phosphate battery state-of-charge monitoring and early warning method based on neural network
CN106897942B (en) Distributed parallel state estimation method and device for power distribution network
CN112630659A (en) Lithium battery SOC estimation method based on improved BP-EKF algorithm
CN105391059A (en) Distributed power generation system state estimation method based on current measurement transformation
Li et al. Estimation of potentials in lithium-ion batteries using machine learning models
CN114430165A (en) Micro-grid group intelligent coordination control method and device based on depth model prediction
CN112909937A (en) Multi-mode digital twinning simulation method and device for rail transit traction power supply system
Lotfi et al. Constraint-guided deep neural network for solving optimal power flow
CN106022970B (en) A kind of active distribution network measure configuration method counted and distributed generation resource influences
Dinavahi et al. ANN based voltage stability margin prediction
Stiasny et al. Solving differential-algebraic equations in power systems dynamics with neural networks and spatial decomposition
CN112700028B (en) Estimation method and system for equivalent inertia and damping space-time distribution of virtual power plant
Chakrabarti Voltage stability monitoring by artificial neural network using a regression-based feature selection method
CN107634536B (en) Method and system for calculating maximum power transmission capacity of alternating current-direct current hybrid system
Morita et al. On‐line detection of state‐of‐charge in lead acid battery using radial basis function neural network
Nissing et al. Thermal model parameter identification of a lithium battery
Sangwan et al. High fidelity equivalent circuit model parameter extraction for characterization and simulation of li-ion cells in battery electric vehicles
JP2022044943A (en) Battery degradation diagnosing system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant